“New vision technology speeds up and improves the rose sorting process.”

Meewisse Roses has overhauled its sorting process. The existing machine, equipped with the IRISS Roses vision system from 4More Technology, has been expanded with additional cameras and AI software for detecting defects. Although IRISS Defects is still learning, the benefits are already clearly noticeable, according to Tom Meewisse and coordinator Katharina Verheul: sorting is significantly faster and more accurate.

Until 2016, roses could only be automatically sorted by length, stem thickness, and bud size. With the introduction of the Intelligent Rose Inspection & Sorting System (IRISS), 4More Technology (4MT.nl) took the sorting process to a higher level of precision, uniformity, and efficiency.

Early last year, the high-tech company from Moerkapelle took another step with the development of IRISS Defects. This is an AI-driven expansion module designed to rapidly detect defects such as bent stems and necks, ball centers (abnormal bud formations that prevent full blooming), and powdery mildew symptoms on the underside of leaves.

“Until recently, our employees had to assess all of this themselves while placing the roses,” says grower Tom Meewisse. “If they detected a defect, they had to set a coding switch above the fork to a specific position to assign the correct downgrade code to the rose, so the machine would send it to the correct bunching station. Experienced workers are quite skilled at this, but it always takes extra time—especially when the flower has to be turned to check whether the underside of the leaves is healthy and undamaged.”

Automatic defect recognition

The Bleiswijk-based grower of the ‘Red Naomi’ rose didn’t have to think long when asked by 4MT whether he was interested in a new development ready for real-world testing. It involved an upgrade of IRISS Roses with a self-learning module for automatically detecting and recognizing defects.

“From our customers, we already knew that defect assessment is a limiting factor in the sorting process,” says Tim van der Elst, owner and operational manager of 4MT. “Tom is open to innovation and understands that AI and its implementation require time and guidance.”

Training the software

Co-owner and head of R&D Wouter Vreugdenhil adds: “You have to train the software to recognize deviations detected by the cameras. Some of these can be variety-specific. For example, Red Naomi is relatively sensitive to ball formation in the center of the flower, and powdery mildew spots on the underside of leaves are somewhat harder to detect. Initially, a lot of manual work is required to teach the software to make the right decisions. For most parameters, enough data has now been collected and entered; for some, we are still in the learning process. It’s still a work in progress, but IRISS Defects is already ready for broader rollout within the rose sector. IRISS Roses and the defects module can be installed on all modern sorting machines, regardless of brand.”

Significantly faster and better

“There is still progress being made in the sorting process,” confirms Meewisse. “It’s not flawless yet, but the speed is much higher and fewer mistakes are made compared to manual assessment. The workers only need to place the roses; they can leave the visual inspection entirely to the vision system. In principle, we could reduce the team from five to four people on the line and still finish sorting earlier than before. We’re done about 1.5 to 2 hours earlier.”

Work coordinator Katharina agrees. “Staffing levels are sometimes already lower, but not yet structurally,” she says. “In any case, the team leader now has more time to manage surrounding tasks. Another advantage is that training new employees takes much less time and they reach full speed more quickly. That also saves time for us as supervisors. And because people differ, while software consistently makes objective decisions based on data history, the uniformity of sorting has also improved. It’s a win-win-win. I’m really happy with it.”

Future outlook

The users are firm in their conclusion that IRISS Defects performs well and is a valuable addition to the 4MT vision system that has been in use for some time. Meewisse hopes that, in the future, data from the sorting process can also be linked to the location in the greenhouse where the rose was harvested.

“We also record a lot in the greenhouse. If you can link stem defects to path numbers and identify patterns, new opportunities for optimization will emerge. You could think of crop protection and scouting, and providing feedback to employees who deliver many roses with leaf or stem damage.”

“We’re not there yet, and other parties will also be needed, but it’s only a matter of time,” Van der Elst concludes. “Technology and AI are increasingly pushing boundaries.”

More information on IRISS Roses, click here